You want to optiooize the optimal transmission path.
First thing you need to do is decide what you want to optimize. It is the total time required to transmit one message from each location to eventually reach a particular host? Is it the total energy required to transmit one message from each location to eventually reach a particular host? Something else? You will need to design your "cost" function to calculate this value.
Next thing you need to do is figure out what your model parameters are.
None of your model parameters should be relative constants: you can use parameterization to pass constants to the cost function. Anything that is to be consistent for the duration of this optimization run should be parameterized. For example if you are trying to optimize costs "right now" for nodes that are temporarily frozen in position, then the current locations could be passed in as parameters.
Your model parameters should be the things that you want to experimentally vary to see which one is the "best" for the current purpose (results in the lowest cost.) And any one collection of model parameters must be complete enough that the calculating the cost associated with those parameters is determininistic.
Suppose for example that you were doing a TSP (Travelling Salesman Problem), then the model parameters could be the nodes in the path, from start until end: knowing the order of nodes and given a set of current constants such as positions of node, you could calculate the "cost" of the TSP path in an unambiguous way.
However, if you are talking about cluster heads, then you do not have a simple path: in this kinds of situations, there is usually a minimum of two hops involved: individual node to its cluster head, and cluster heads to the host. What would each entry in the model parameters represent?
Well, in the case where there are exactly two hops for regular nodes (first to cluster head, then cluster head to host) and one hop for each of the nodes that are already cluster heads, and there will not be any further hops (no relaying along chains of nodes), then what your model parameters could be, is which node number to use a cluster head for each node. The cost calculation would then run through and determine the cost to get from the node to the designated cluster head, and then the cost from that cluster head to the host. Any node that nominated itself as its own cluster head would have cost 0 to reach itself in the first half of the calculation.
The above calculation would have the possibility that some nodes might be "informal" cluster heads, used only by a small number of nodes, because any node could pick any other node to send to first. But if you instead wanted to designate particular cluster heads, then you would create a table of cluster head information first, then restrict each model parameter to be an integer 1 to the number of already-designated cluster heads.
